Jev 1.14

jev-1.14

Jev 1.14 is TypeSafe's System One model for fast, structured decision-making in software. It evaluates text and application state against developer-defined questions, returning choices, yes/no probabilities, and scores that applications can use directly. Built for classification, routing, and automated workflows, it helps developers turn natural-language judgments into actionable outputs and supports evaluating multiple questions in parallel within a single request.

Context Window

32K tokens

Release Date

Oct 14, 2026

Modalities

Jev 1.14 Pricing

Input PriceOutput Price
$0.042/M$0/M

Jev 1.14 API Capabilities

Reasoning

Not supported

Tool calling

Not supported

Temperature parameter

Not supported

Attachments

Not supported

Knowledge Base

—

Endpoint Protocols

systemone

Jev 1.14 — Model Highlights

Verified strengths and model characteristics.

Typed Probabilistic Decisions

Returns choices, rubric scores and yes/no probabilities for focused judgments that application code can combine.

Parallel State Evaluation

Evaluates independent questions against shared context in parallel, reducing repeated processing in decision workflows.

Experimental Visual Grounding

Adds a Box question type that locates one image target with normalized coordinates and confidence; image features remain experimental.

Jev 1.14 — Use Cases

Practical workloads supported by official model information.

Support Intent Triage

Classify a support request, assess urgency and route it to a handler or reviewer using application-defined thresholds.

Retrieval Passage Scoring

Score retrieved passages for relevance before selecting evidence for a separate answer-generating model.

Image Crop Prototyping

Prototype semantic image crops by checking target presence and locating it with Box; review results before production use.

How to Use Jev 1.14 via the TokenHub API

Create API key
curl 'https://us-api.tokenhub.com/v1/systemone' \
  -X 'POST' \
  -H "Authorization: Bearer $TOKENHUB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "assets": {
    "a7f291ce-test-photo": {
      "data_base64": "<BASE64_JPEG_BYTES>",
      "format": "jpeg",
      "type": "image"
    }
  },
  "model": "jev-1.14",
  "questions": {
    "department": {
      "criteria": {
        "billing": "Charges and refunds",
        "technical": "Product defects"
      },
      "instructions": "Which team should handle this ticket?",
      "type": "choice"
    },
    "needs_refund": {
      "instructions": "Does the customer explicitly request a refund?",
      "type": "noul"
    },
    "person_box": {
      "instructions": "the person in `photo`",
      "type": "box"
    },
    "urgency": {
      "criteria": [
        "Routine",
        "Needs attention",
        "Critical"
      ],
      "instructions": "How urgent is this ticket?",
      "type": "score"
    }
  },
  "state": {
    "photo": "a7f291ce-test-photo",
    "ticket": "I was charged twice. Please refund the duplicate charge."
  }
}'

Jev 1.14 — Media and Demos

Verified public posts and videos; community findings are not guarantees.

X (Twitter)

View original
View post on X
View original
View post on X
View original
View post on X

Reddit

View original
View original
View original

YouTube

View original
Watch on YouTube
View original
Watch on YouTube
View original
Watch on YouTube

Jev 1.14 — FAQs

Model selection, limitations, pricing and TokenHub integration.

What Is Jev 1.14?+

Jev 1.14 is TypeSafe AI’s System One decision model update, released as jev-1.14.0 through jev-preview. It produces typed judgments for software.

When Should I Choose Jev 1.14?+

Choose it for focused routing, classification and scoring with predefined answers. Use a generative model for prose, code or extended reasoning explanations.

Does Calibration Guarantee Correct Answers?+

No. Calibration describes groups of predictions, not guaranteed correctness for one decision. Validate thresholds on representative data, including difficult and non-English inputs.

What Are Jev 1.14’s Experimental Limits?+

Images and Box are experimental. Box returns coordinates even when the target is absent. Default context is 64K per request and 32K per question; opt-in 262K context is slower and less accurate.

How Do I Use Jev 1.14 On TokenHub?+

If a route is available, follow TokenHub’s exact model ID and decision-request example. Confirm endpoint, question types and experimental-header support; do not assume standard chat request semantics.

How Is Jev 1.14 Priced?+

TypeSafe lists $0.042 per million input tokens and no output-token charge. Check TokenHub’s current route pricing and availability separately.

Ready to use Jev 1.14?

Use one API key to access Jev 1.14 and more AI models through TokenHub.

Create API key